2012-08-21 16:54:53

by Seiji Aguchi

[permalink] [raw]
Subject: [RFC][PATCH v3 3/3] efi_pstore: Skiping scheduling a workqueue in cases other than oops

[Problem]
efi_pstore creates sysfs files when logging kernel messages to NVRAM.
Currently, the sysfs files are updated in a workqueue which is registered in a write callback.

On the other hand, situations which users needs the sysfs files are when they erase entries or oops happen
because a system will be down and users can't access to sysfs files in other cases like panic, reboot or emergency_restart.

Also, if kernel panics due to a bug of workqueue operations and a write callback of efi_pstore is called in
panic case, efi_pstore may fail due to a failure of schedule_work().
And panic_notifier_chain()/emergency_restart() is not kicked if efi_pstore fails.
This may cause user's unwanted result.

[Patch Description]
This patch registers a workqueue updating sysfs entries in cases where users erase entries or oops happen only,
and skips it in other cases like panic, reboot or emergency_start.

Signed-off-by: Seiji Aguchi <[email protected]>
---
drivers/firmware/efivars.c | 7 ++++++-
1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
index 3dde7d6..a4919ad 100644
--- a/drivers/firmware/efivars.c
+++ b/drivers/firmware/efivars.c
@@ -746,7 +746,12 @@ static int efi_pstore_write(enum pstore_type_id type,

spin_unlock_irqrestore(&efivars->lock, flags);

- schedule_work(&efivar_work);
+ /*
+ * The user may want to update sysfs for this write
+ * when they erase an entry via /dev/pstore or oops happen.
+ */
+ if (!size || reason == KMSG_DUMP_OOPS)
+ schedule_work(&efivar_work);

*id = part;
return ret;
-- 1.7.1